How Non-Human Workers Struggle on Real Streets
At the Fortune Brainstorm AI conference on December 26, 2025, Serve Robotics co-founder MJ Burk Chun revealed surprising challenges faced by autonomous delivery robots — what she described as strikingly human-like reactions when navigating unpredictable public environments. Although we often picture robots as precise and logical, the reality for these AI employees is far more complex. Once deployed outside simulations into real cities like Los Angeles, robots encounter unusual obstacles — from aggressive cars to unexpected creatures — that their training data didn’t prepare them for. For example, one robot literally froze when confronted with a baby goat on the sidewalk, an outlier that threw off its perception system and halted its movement.
Algorithms vs. The Real World
Chun explained that robots must constantly assess risk and adapt to dynamic conditions. Intersections filled with speeding vehicles generate what she called robot “nightmares” about cars — not literal dreams, but constant calculations about survival and safety before crossing a street. This reveals a major gap between controlled algorithmic training and the messy unpredictability of real human environments. Serve Robotics’ delivery machines are thus much more than simple courier bots: they’re constantly learning and recalibrating their understanding of safety, public behavior, and physical space.
From LA to Miami Beach: Culture Shock for Robots
In another striking example, robots experienced what Chun likened to culture shock when Serve expanded its fleet from Los Angeles to Miami Beach. The routing algorithms designed for the fast-paced, high-traffic intersections of LA didn’t mesh with how drivers and pedestrians move in Florida’s more relaxed traffic flow. This mismatch meant non-human workers had to be re-tuned to local conditions — a problem that mirrors how humans adapt to new cities. The lesson? Deploying autonomous technologies nationally (or globally) requires deep consideration of local movement patterns and social norms, not just software calibration.
Robots as Social Ambassadors
Finally, Chun emphasized that robots on sidewalks don’t have inherent rights there — people do. Serve Robotics engineers therefore design systems that prioritize human comfort and social acceptance over raw robotic efficiency. These robots must “deliver delight” to communities, and in doing so are evolving into multifunctional tools. Beyond deliveries, they now gather data on missing curb cuts and hidden potholes to help cities improve infrastructure — a role far broader than originally intended.
Key Highlights:
- Real-World AI Challenges: Robots encounter untrained scenarios like baby goats and aggressive traffic, showing limits of current AI training.
- Adaptation Required: Algorithms must adjust to different city behaviors, showing robots can face culture shock.
- Social Integration: Robots are treated as ambassadors in public spaces, prioritizing human-centric design.
Reference:
